Skip to main content
Diabetes, Metabolic Syndrome and Obesity logoLink to Diabetes, Metabolic Syndrome and Obesity
. 2026 Sep 30;19:620529. doi: 10.2147/DMSO.S620529

Body Mass Index and Cerebral Microbleeds in Type 2 Diabetes: Sex-Specific Thresholds from an Exploratory Study

Jian Yuan 1,*, Yun Deng 1,*, Xingyu Chen 1, Feihong Chen 2, Linrui Wang 3, Ning Rong 4,✉, Desheng Zhu 1,5,✉
PMCID: PMC13635289  PMID: 42835389

Abstract

Objective

To prospectively investigate the association between baseline body mass index (BMI) and the risk of incident cerebral microbleeds (CMBs) in middle-aged and elderly patients with type 2 diabetes mellitus (T2DM), and to explore potential gender-specific risk thresholds in an exploratory analysis.

Methods

This single-center prospective observational cohort study consecutively enrolled T2DM patients without baseline CMBs from April 2022 to March 2023. Baseline demographic, clinical, and biochemical data were collected. BMI was categorized according to Chinese criteria (normal weight BMI < 24 kg/m2, overweight 24–28 kg/m2, and obesity BMI ≥ 28 kg/m2). All participants underwent follow-up 3.0T magnetic resonance imaging (including susceptibility-weighted imaging sequence) every 12–18 months or earlier if clinically indicated to detect incident CMBs. The primary outcome was the occurrence of ≥1 incident CMBs during follow-up. Cox proportional hazards models were used to calculate hazard ratios (HRs). Restricted cubic spline (RCS) models and two-piecewise linear regression were employed to analyze the nonlinear relationship and threshold effects between BMI and CMBs risk.

Results

The study ultimately included 936 T2DM patients without baseline CMBs, with a median follow-up of 30.00 months (range: 6–42 months; IQR: 12.0, 36.0 months). During follow-up, 100 patients (10.68%) developed incident CMBs. Multivariable Cox regression analysis revealed that compared with the normal weight group, the obesity group (BMI ≥ 28 kg/m2) had a significantly higher risk of incident CMBs (HR = 2.58, 95% CI: 1.46–4.54, p < 0.001). Each 1 kg/m2 increase in baseline BMI was associated with a 15% increased risk of incident CMBs (HR = 1.15, 95% CI: 1.09–1.22, p < 0.001). Threshold effect analysis showed a nonlinear association between BMI and CMBs risk (p for nonlinearity < 0.001). The exploratory inflection point was 21.10 kg/m2 for males; beyond this threshold, each unit increase in BMI increased CMBs risk by 18% (HR = 1.18, 95% CI: 1.10–1.26, p < 0.001). For females, the exploratory inflection point was 26.80 kg/m2, beyond which risk increased by 25% per unit increase (HR = 1.25, 95% CI: 1.02–1.54, p = 0.029).

Conclusion

In this single-center Chinese cohort, higher baseline BMI was associated with incident CMBs in middle-aged and elderly T2DM patients. Exploratory analysis suggested potential sex-specific inflection points (male: 21.10 kg/m2; female: 26.80 kg/m2). This study may inform future research on gender-differentiated weight management strategies, but these findings are not yet ready for clinical implementation.

Keywords: diabetes mellitus, body mass index, cerebral microbleeds, prospective study, threshold effect

Graphical Abstract

A graphical abstract on BMI and cerebral microbleeds in T2DM, showing cohort, BMI exposure and outcome details. The graphical abstract titled ′BMI and Cerebral Microbleeds in T2DM: Sex-Specific Thresholds′ presents three main sections: Cohort, BMI exposure and Outcome. The Cohort section describes 936 T2DM patients with no baseline CMBs. The BMI exposure section indicates an increase of 1 kg/m superscript 2, with HR equals 1.15 (1.09 to 1.22) and p less than 0.001. Obesity (BMI greater than or equal to 28) has HR 2.58. The Outcome section shows 100 incident CMBs, accounting for 10.7 percent of the cohort, with deep CMBs having HR 1.21. Exploratory sex-specific thresholds are noted: for males, 21.10 kg/m superscript 2 (greater than or equal to 21.10: HR 1.18) and for females, 26.80 kg/m superscript 2 (greater than or equal to 26.80: HR 1.25). The abstract concludes with ′Higher BMI associated with incident CMBs, nonlinear threshold, sex dimorphism.′.

Introduction

Cerebral microbleeds (CMBs) are key imaging biomarkers of cerebral small vessel disease, appearing as 2–10 mm ovoid hypointense foci on susceptibility-weighted imaging (SWI) or gradient-recalled echo (GRE) magnetic resonance imaging (MRI) sequences.1 The prevalence of CMBs ranges from 5% to 23% in the general elderly population, while in patients with type 2 diabetes mellitus (T2DM), this proportion can be as high as 20–30%.2 CMBs are not only independent risk factors for future symptomatic intracerebral hemorrhage and ischemic stroke3 but are also closely associated with cognitive decline, gait disturbances, and vascular dementia,4 significantly impacting the quality of life and prognosis of elderly patients.

Body mass index (BMI), as a simple indicator of overall obesity, exhibits a complex relationship with cerebrovascular disease. The classic “obesity paradox” suggests that overweight or mild obesity may be associated with lower mortality in specific populations (eg, elderly, chronic disease patients).5 However, through multiple pathophysiological pathways including insulin resistance, chronic low-grade inflammation, endothelial dysfunction, increased oxidative stress, and sympathetic activation, obesity is undoubtedly a significant driver of atherosclerosis and small vessel disease.6 This paradoxical phenomenon may partly stem from analyzing BMI as a linear variable, overlooking potential nonlinear “J-shaped” or “U-shaped” associations with clinical outcomes, and insufficiently considering modifying effects of gender, age, and underlying disease status.

Current research on the association between BMI and CMBs primarily employs cross-sectional designs,7,8 making it difficult to establish causal temporal relationships. A few prospective studies have mainly focused on general populations or stroke patients,9 with particularly scarce prospective evidence specifically targeting the high-risk T2DM population. Furthermore, previous studies typically treat BMI as a continuous or simple categorical variable, potentially masking the true dose-response relationship shape and critical “risk inflection points”.

Therefore, given the exploratory nature of sex-specific threshold analysis in this field, this study aimed to: 1) prospectively evaluate the association between baseline BMI levels and incident CMBs risk in middle-aged and elderly T2DM patients; and 2) explore potential nonlinear associations and generate hypotheses regarding gender-specific risk thresholds that may warrant validation in future studies.

Methods

Study Registration and Ethics

This study was approved by the Ethics Committee of Baoshan Branch, Renji Hospital, Shanghai Jiao Tong University School of Medicine (Approval No.: 2021-qkt-005). As a prospective observational study with a non-interventional design, it was not registered as a clinical trial. All participants provided written informed consent after being fully informed of the study purpose, procedures, and potential risks.

Study Design

This single-center, prospective, observational cohort study aimed to investigate the association between baseline BMI and incident CMBs in middle-aged and elderly T2DM patients, and to explore gender-specific risk thresholds. Participants were recruited from the neurology and endocrinology outpatient clinics or inpatient departments of our hospital from April 1, 2022 to March 31, 2023. All subjects had no CMBs at baseline and underwent regular MRI follow-ups to detect incident events.

Sample Size Estimation

As an exploratory cohort study, the sample size was estimated based on feasibility and event rates. Assuming an annual CMBs incidence of approximately 3–5% in middle-aged and elderly T2DM patients, with a planned 3-year follow-up, the expected cumulative incidence was approximately 10%. To detect a 15% increased risk per 1 kg/m2 increase in BMI (HR=1.15), with α=0.05 and β=0.20, PASS software estimated a required sample size of approximately 900, consistent with the final included sample size (n=936).

Study Population

The study population consisted of T2DM patients who visited or were hospitalized at the dizziness or stroke outpatient clinic of the Department of Neurology, Baoshan Branch, Renji Hospital, Shanghai Jiao Tong University School of Medicine, from April 1, 2022 to March 31, 2023.

Inclusion Criteria: 1). Age 47–86 years; 2). Meeting the World Health Organization (WHO) 2019 diagnostic criteria for diabetes;10 3). Baseline 3.0T head MRI (including SWI sequence) confirming absence of CMBs; 4). Life expectancy > 3 years, as determined by the attending neurologist based on clinical assessment incorporating age, Charlson Comorbidity Index score, and overall functional status, with agreement to complete regular follow-up.

Exclusion Criteria: 1). Presence of other significant brain structural lesions (eg, brain tumors, large infarctions, hemorrhages, arteriovenous malformations) on baseline MRI; 2). Evidence or history of clinically diagnosed probable cerebral amyloid angiopathy (CAA) (eg, lobar hemorrhage based on Boston criteria); 3). Severe organ dysfunction (NYHA class III–IV heart failure, Child-Pugh B/C liver function, estimated glomerular filtration rate (eGFR) < 30 mL/min/1.73m2); 4). Comorbid hematologic diseases, immune system disorders, or active malignancies; 5). Contraindications to MRI (eg, cardiac pacemaker, claustrophobia); 6). Pregnant or lactating women.

Baseline Data Collection

During the baseline visit, the following information was collected through standardized questionnaires, physical examinations, and medical record reviews: 1). Demographics and lifestyle: age, gender, education level, smoking history (defined as current or former regular smoking with total cigarette consumption ≥ 100), alcohol consumption history (defined as regular drinking ≥ 1 year, with weekly alcohol intake equivalent to ≥ 100 grams of pure ethanol). 2). Medical history and medication use: duration of diabetes (calculated from diagnosis to the incident CMBs or censoring), hypertension history, hyperlipidemia history, coronary heart disease history, stroke history, and use of antidiabetic, antihypertensive, statin, and antiplatelet medications. 3). Physical measurements: height, weight, and blood pressure [diastolic blood pressure (DBP), and systolic blood pressure (SBP)] were measured by trained nurses using standardized equipment. Blood pressure was taken as the average of three consecutive measurements after 5 minutes of rest. BMI was calculated as weight (kg) divided by height squared (m2). 4). Laboratory Tests: fasting venous blood was collected to measure glycated hemoglobin (HbA1c), fasting blood glucose (FBG), lipid profile [total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C)], liver function [alanine aminotransferase (ALT), aspartate aminotransferase (AST)], and renal function [serum creatinine, uric acid (UA) and eGFR].

Exposure Definition and Grouping

The primary exposure variable was baseline BMI. According to the “Guidelines for Prevention and Control of Overweight and Obesity in Chinese Adults”,11 participants were categorized into three groups: normal weight (BMI < 24 kg/m2), overweight (24 ≤ BMI < 28 kg/m2), and obesity (BMI ≥ 28 kg/m2). BMI was analyzed both as a continuous variable (per 1 kg/m2 increase) and using these three predefined categories. For exploratory supplementary analyses, the WHO-recommended Asian-population obesity threshold (BMI ≥ 25 kg/m2)12 was additionally applied for descriptive binary grouping, with results presented in the supplementary materials. For hierarchical analysis, continuous stratifying variables were dichotomized into two groups based on their respective median values.

Outcome Assessment

Primary Outcome: Incident CMBs during follow-up, defined as the appearance of at least one new lesion on follow-up MRI SWI sequences that was not present on baseline MRI, meeting imaging criteria for CMBs (2–10 mm homogeneous hypointense foci, excluding vascular flow voids, calcifications, and iron deposits). MRI Protocol and Interpretation: All participants underwent 3.0T head MRI (GE Signa Premier) at baseline and every 12–18 months thereafter, including T1-weighted imaging, T2-weighted imaging, fluid attenuated inversion recovery, diffusion-weighted imaging, and SWI sequences. SWI parameters: TR 27 ms, TE 20 ms, flip angle 15°, slice thickness 1.0 mm, matrix 384×384. All MRI images were independently interpreted by two experienced neuroradiologists blinded to clinical data. Disagreements regarding the presence or novelty of CMBs were resolved by a third senior expert.

Follow-Up Procedures

Follow-up was conducted through outpatient reviews, telephone calls, and medical record system tracking. Follow-up time started from the baseline MRI date and ended at the occurrence of the primary outcome (incident CMBs), death, loss to follow-up, or the study cutoff date (October 31, 2025), whichever came first. While the scheduled MRI follow-up interval was every 12–18 months, the actual time to the primary outcome was recorded based on the date of the first MRI scan that detected incident CMBs. For patients who underwent unscheduled neuroimaging due to new-onset neurological symptoms (eg, stroke or severe dizziness) prior to the 12-month mark, the actual date of these scans was used to calculate the follow-up duration, resulting in a minimum observation period of 6 months for some participants.

Statistical Analysis

All statistical analyses were conducted using R software (version 4.3.2). Continuous variables are presented as mean ± standard deviation or median (interquartile range) according to their distribution, with between-group comparisons performed using Student’s t-test, one-way ANOVA, or nonparametric tests as appropriate. Categorical variables are expressed as number (percentage) and were compared using the χ2 test or Fisher’s exact test.

Event-free survival across BMI categories was visualized using Kaplan-Meier curves, and between-group differences were assessed with the log-rank test. Univariate and multivariate Cox proportional hazards regression models were employed to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between baseline BMI - analyzed both as a continuous and a categorical variable - and the risk of incident CMBs. The multivariate model was constructed by initially including variables with p < 0.05 in univariate analysis, then refining based on clinical relevance and prior literature. Duration of diabetes and HbA1c were included as mandatory confounders account for the cumulative effect of chronic glycemic exposure on cerebral microvascular integrity. The final adjusted model (Model 3) included age, sex, DBP, SBP, duration of diabetes, drinking history, Hb, PLT, AST, TG level, HbA1c, and use of antiplatelet medication. The proportional hazards assumption was verified using Schoenfeld residual tests (all p > 0.05), and no material violations were detected.

To examine potential nonlinearity in the BMI–CMBs relationship, we fitted restricted cubic spline (RCS) models with BMI as a continuous predictor and three knots placed at the 10th, 50th, and 90th percentiles. Nonlinearity was evaluated using the likelihood ratio test; a p-value < 0.05 indicated a significant departure from linearity. Where nonlinearity was present, a two-piecewise linear regression model was applied to identify the inflection point (exploratory analysis), defined as the BMI value that maximized the log-likelihood function. Threshold analyses were stratified by sex.

Missing data were minimal (<5% overall) and largely confined to laboratory measures. We applied multiple imputation with predictive mean matching (m = 5 imputed datasets), including all analysis covariates and the outcome variable. All statistical tests were two-sided, and p < 0.05 was considered statistically significant. In time-to-event analyses, participants who died or were lost to follow-up were censored at their last known contact date. No adjustment for multiple comparisons was made for subgroup analyses, which were considered exploratory in nature.

Results

Follow-Up Completion and Baseline Characteristics of Study Population

The CONSORT flow diagram illustrates participant screening, enrollment, follow-up, and analysis (enrollment flowchart in Figure 1). A total of 1023 patients were enrolled between April 1, 2022 and March 31, 2023. During follow-up (maximum 42 months), 87 (8.50%) were lost to follow-up before reaching the study endpoint. The median time to loss to follow-up was 18.5 months (IQR 9.0–27.0). The primary reasons for loss to follow-up included: death (n=18, 1.76%), refusal to continue participation (n=36, 3.52%), loss of contact (n=25, 2.44%), and other reasons (n=8, 0.78%).

Figure 1.

A flowchart of cohort enrollment, follow-up and analysis for 1023 participants. The flowchart illustrates the process of cohort enrollment, follow-up and analysis. Initially, 1023 participants were enrolled. Out of these, 87 participants (8.5 percent) were lost to follow-up due to reasons such as refusal (36 participants, 3.5 percent), loss of contact (25 participants, 2.4 percent), death (18 participants, 1.8 percent) and other reasons (8 participants, 0.8 percent). The remaining 936 participants (91.5 percent) completed the follow-up. Among those who completed follow-up, 100 participants (9.8 percent) had incident CMBs detected and were included in the primary analysis. The other 836 participants (81.7 percent) had no incident CMBs and were also included in the primary analysis. The 87 participants who were lost to follow-up were censored in the survival analysis.

Study population screening and follow-up flowchart. Flowchart showing patient screening, inclusion, exclusion, and follow-up process.

The study ultimately included 936 T2DM patients without baseline CMBs, enrolled from April 1, 2022 to March 31, 2023. The actual follow-up duration ranged from 6 to 42 months, with a median of 30.0 months (range: 6–42 months; IQR: 12.0, 36.0 months). The first scheduled imaging assessment typically occurring between 12 and 18 months. Based on baseline BMI categories: normal weight group 292 (31.20%), overweight group 404 (43.16%), obesity group 240 (25.64%). Baseline characteristics of the three groups are shown in Table 1. Overweight group patients were slightly younger (p = 0.01) with the highest male proportion (p < 0.01). The obesity group had significantly higher diastolic blood pressure, systolic blood pressure, hypertension prevalence, hyperlipidemia prevalence, hyperuricemia prevalence, AST, UA, TG, and antiplatelet treatment compared to normal weight and overweight groups (all p < 0.05). No significant differences were found between groups in duration of diabetes, smoking, RBC, ALB, PLT, TC, LDL-C, or smoking history. Baseline characteristics by CMBs group are presented in Supplement Table 1, comparing demographic, clinical, and laboratory features between groups with and without CMBs.

Table 1.

Baseline Characteristics of Participants Stratified by BMI Group

Characteristics Total (n = 936) Normal Weight Group (BMI<24) Overweight Group (24≤BMI<28) Obesity Group (BMI≥28) P
Demographic characters
N 936 292 404 240
Age 62.65 ± 7.00 63.61 ± 7.05 62.05 ± 6.92 62.49 ± 6.98 0.01
Sex, n (%) <0.01
Female 402 (42.95) 131 (44.86) 152 (37.62) 119 (49.58)
Male 534 (57.05) 161 (55.14) 252 (62.38) 121 (50.42)
Clinical features
DBP (mmHg) 78.25 ± 11.83 74.91 ± 11.47 78.17 ± 11.40 82.46 ± 11.68 <0.01
SBP (mmHg) 137.57 ± 19.68 133.78 ± 19.02 137.39 ± 19.51 142.49 ± 19.77 <0.01
BMI (kg/m2) 25.85 ± 3.69 21.98 ± 1.53 25.79 ± 1.15 30.66 ± 2.57 <0.01
Duration of Diabetes (years) 21.00 (7.00–34.00) 23.00 (6.00–35.00) 22.00 (7.00–34.00) 18.00 (6.00–33.00) 0.38
Medical history, n (%)
Smoking 451 (48.18) 135 (46.23) 209 (51.73) 107 (44.58) 0.15
Drinking 218 (23.29) 54 (18.49) 110 (27.23) 54 (22.50) 0.03
Hypertension 433 (46.26) 100 (34.25) 193 (47.77) 140 (58.33) <0.01
Hyperlipidemia 140 (14.96) 17 (5.82) 70 (17.33) 53 (22.08) <0.01
Hyperuricemia 28 (2.99) 4 (1.37) 5 (1.24) 19 (7.92) <0.01
Laboratory data
WBC, (×109/L) 6.79 ± 2.92 6.57 ± 3.12 6.73 ± 2.09 7.15 ± 3.75 0.06
RBC, (×1012/L) 4.68 ± 0.60 4.65 ± 0.70 4.68 ± 0.55 4.69 ± 0.55 0.71
Hb, (g/L) 140.93 ± 16.12 138.58 ± 15.33 142.69 ± 16.67 140.81 ± 15.79 <0.01
PLT, (×109/L) 215.71 ± 64.49 216.45 ± 64.54 214.80 ± 65.39 216.35 ± 63.12 0.93
AST, (U/L) 20.00 (16.00–26.00) 18.00 (15.00–22.00) 19.00 (15.00–26.00) 23.00 (19.00–33.00) <0.01
ALB, (g/L) 42.13 ± 6.88 42.30 ± 7.39 41.88 ± 6.23 42.35 ± 7.28 0.61
UA, (μmol/L) 337.02 ± 97.24 310.43 ± 82.82 341.31 ± 108.16 362.03 ± 85.92 <0.01
TC, (mmol/L) 4.82 ± 1.35 4.75 ± 1.28 4.91 ± 1.43 4.75 ± 1.29 0.22
HDL, (mmol/L) 1.18 ± 0.47 1.29 ± 0.62 1.12 ± 0.35 1.13 ± 0.44 <0.01
LDL, (mmol/L) 3.13 ± 0.90 3.07 ± 0.91 3.17 ± 0.91 3.11 ± 0.88 0.34
TG, (mmol/L) 1.60 (1.14–2.50) 1.33 (0.92–1.97) 1.71 (1.25–2.54) 2.04 (1.40–3.01) <0.01
HbA1c, (%) 9.35 ± 2.30 9.46 ± 2.46 9.48 ± 2.25 9.00 ± 2.13 0.02
Treatment during follow-up
Hypoglycemic drug 693 (74.20) 232 (79.45) 297 (73.51) 164 (68.33) <0.01
Lipid-lowering drug 95 (10.15) 31 (10.62) 41 (10.15) 23 (9.58) 0.93
Antiplatelet drug 454 (48.50) 150 (51.37) 189 (46.78) 115 (47.92) 0.48
Antihypertensive drug 317 (33.87) 72 (24.66) 140 (34.65) 105 (43.75) <0.01
Follow-up outcomes
Follow-up duration (months) 30.00 (12.00, 36.00) 30.00 (18.00,36.00) 24.00 (12.00,36.00) 30.00 (12.00,36.00) 0.12
CMBs, n (%) 100 (10.68) 20 (6.85) 28 (6.93) 52 (21.67) <0.01

Abbreviations: DBP, Diastolic blood pressure; SBP, Systolic blood pressure; BMI, Body mass index; WBC, White blood cell; RBC, Red blood cell; Hb, Hemoglobin; PLT, Platelet; AST, Aspartate aminotransferase; ALB, Albumin; UA, Uric acid; TC, Total cholesterol; HDL, High-density lipoprotein; LDL, Low-density lipoprotein; TG, Triglycerides; HbA1c, Glycated hemoglobin; CMBs, Cerebral Microbleeds.

Incidence of CMBs During Follow-Up

During the follow-up, 100 patients (10.68%) developed incident CMBs. By BMI category, CMBs incidence increased with higher BMI: normal weight group 20 (6.85%), overweight group 28 (6.93%), obesity group 52 (21.67%), with trend test p < 0.001, consistent when stratified by gender (Supplement Table 2). Kaplan-Meier survival curves (Figure 2) showed significant differences in CMBs-free survival among the three groups (Log-rank test p < 0.001), with the obesity group having significantly higher cumulative risk than the other two groups.

Figure 2.

A Kaplan Meier line graph showing non CMBs probability over follow up time by body mass index group. Kaplan Meier graph with BMI categories: Normal weight, Overweight, Obesity. Vertical axis: Non CMBs probability (0.6-1.0). Horizontal axis: Follow-up time (0-48 months). Normal weight probabilities: (0, 1.0), (6, 1.0), (12, 0.98), (18, 0.97), (24, 0.94), (30, 0.93), (36, 0.92), (42, 0.87). Overweight probabilities: (0, 1.0), (6, 1.0), (12, 0.98), (18, 0.96), (24, 0.94), (30, 0.93), (36, 0.91), (42, 0.80). Obesity probabilities: (0, 1.0), (6, 1.0), (12, 0.95), (18, 0.94), (24, 0.88), (30, 0.85), (36, 0.69). Risk numbers at times 0, 6, 12, 18, 24, 30, 36, 42, 48: Row 1: 292, 292, 284, 220, 196, 156, 126, 38, 0. Row 2: 404, 404, 392, 284, 255, 200, 142, 34, 0. Row 3: 240, 240, 230, 179, 170, 126, 93, 21, 0.

Kaplan–Meier survival curves for CMBs-free survival according to baseline BMI categories. The cumulative incidence-free survival probability is presented for three BMI groups: normal weight (BMI < 24 kg/m2; red solid line), overweight (24 ≤ BMI < 28 kg/m2; green dashed line), and obesity (BMI ≥ 28 kg/m2; blue dotted line). The number at risk during follow-up for each group is shown below the curve. The log-rank test demonstrated a statistically significant difference in survival distributions among the three groups (p < 0.001).

Cox Proportional Hazards Regression Analysis

Univariate Cox regression analysis and confounder selection. In univariate Cox regression analysis (Supplement Table 3), incident CMBs risk was significantly associated with DBP (HR=1.05, 95% CI: 1.03–1.06), SBP (HR=1.02, 1.01–1.03), BMI (HR=1.18, 1.13–1.24), drinking history (HR=1.82, 1.20–2.76), Hb (HR=1.03, 1.02–1.05), PLT (HR=1.01, 1.00–1.01), AST (HR=1.01, 1.01–1.02), TG (HR=1.09, 1.02–1.15), and antiplatelet drug use (HR=1.55, 1.04–2.30) (all p < 0.05). Duration of diabetes and HbA1c were included as mandatory confounders in the final adjusted models, regardless of its univariate p-value > 0.05, to account for the cumulative effect of chronic glycemic exposure on cerebral microvascular integrity.

Multivariate Cox Regression Analysis. In the unadjusted model (Model 1), compared with the normal weight group, the obesity group had a 3.52-fold increased risk of incident CMBs (HR = 3.52, 95% CI: 2.10–5.91, p < 0.001), while the overweight group showed no significant risk increase. After stepwise adjustment for potential confounders including sex, age, DBP, SBP, duration of diabetes, drinking history, Hb, PLT, AST, TG, HbA1c, and antiplatelet drugs (Model 3), the hazard ratio for the obesity group attenuated but remained statistically significant, HR = 2.58 (95% CI: 1.46–4.54, p < 0.001). When analyzing BMI as a continuous variable, in the fully adjusted model, each 1 kg/m2 increase in baseline BMI was associated with a 15% increased risk of incident CMBs (HR = 1.15, 95% CI: 1.09–1.22, p < 0.001). Results are shown in Table 2. Exploratory Supplementary analysis adopting the WHO Asian BMI cutoff (≥25 kg/m2) showed that the higher BMI group had a significantly increased risk of incident CMBs (adjusted HR: 2.24; 95% CI: 1.41–3.55). Detailed results are provided in the Supplement Table 4.

Table 2.

Multivariate Cox Regression Analysis for Effects of BMI Index on Risk of CMBs

Patients, N Model 1 (Unadjusted) Model 2 (Adjusted) Model 3 (Adjusted)
Incidence, n (%) HR (95% CI) P N HR (95% CI) P N HR (95% CI) P
BMI index
(continuous, for every 1 kg/m2 increase), 936
100 (10.68) 1.18 (1.11, 1.25) < 0.001 936 1.16 (1.09, 1.23) < 0.001 934 1.15 (1.09, 1.22) < 0.001
BMI index (categorical)
Normal weight group (15.00–23.90), 292 20 (6.85) 1.0 (reference) 292 1.0 (reference) 291 1.0 (reference)
Overweight group (24.00–27.90), 404 28 (6.93) 1.11 (0.62, 1.97) 0.722 404 1.12 (0.63, 2.00) 0.692 404 1.01 (0.55, 1.86) 0.973
Obesity group (28.00–40.70), 240 52 (21.67) 3.52 (2.10, 5.91) <0.001 240 3.65 (2.17, 6.12) <0.001 239 2.58 (1.46, 4.54) <0.001
P for trend < 0.01 < 0.01 < 0.01

Notes: Model 1: unadjusted; Model 2: adjusted for sex, age; Model 3: adjusted for sex, age, DBP, SBP, duration of diabetes, drinking history, Hb, PLT, AST, TG, HbA1c, and antiplatelet drugs (Covariates with a p-value < 0.05 in the univariate analysis were included, and duration of diabetes and HbA1c were included as a mandatory clinical variable).

Subgroup, Sensitivity, and Anatomical Location-Stratified Analyses

To test the stability of our findings, subgroup and sensitivity analyses were performed. Subgroup analysis (Figure 3 and Supplement Table 5) showed that the positive association between BMI and CMBs risk remained consistent across all predefined subgroups (all interaction p > 0.05). Sensitivity analyses further confirmed these results (Supplement Table 6), with the association persisting after excluding patients with smoking, hypertension, hyperlipidemia, or hyperuricemia, and across strata defined by lipid-lowering/antihypertensive drug use and medication stability. When stratified by anatomical location, the association was stronger for deep CMBs (HR = 1.21, 95% CI: 1.10–1.33) than for lobar CMBs (HR = 1.09, 95% CI: 0.98–1.22), suggesting that obesity-related vasculopathy may predominantly affect deep perforating arteries.

Figure 3.

A forest plot of body mass index increase versus incident cerebral microbleeds across subgroups. Forest plot with columns labeled, Hierarchical factor, N, HR (95 percentCI) P and P for interaction. A dashed vertical reference line marks hazard ratio equals 1.0. The x-axis is labeled, Risk for CMBs, ranging from 0.9 to 1.5 with ticks at 0.9, 1.0, 1.1, 1.2, 1.3, 1.4 and 1.5. The y-axis lists subgroup rows: Sex, Age, DBP, SBP, Duration of diabetes, Drinking history, Hb, PLT, AST, TG, HbA1c and Antiplatelet drugs. Each row shows a square point estimate with a horizontal 95 percent confidence interval. Values shown: Male N 533 hazard ratio 1.17 (1.09, 1.24) P less than 0.001; Female N 401 hazard ratio 1.13 (1.05, 1.23) P 0.002; P for interaction 0.563. Age less than 62 years N 422 hazard ratio 1.19 (1.06, 1.33) P 0.002; Age greater than or equal to 62 years N 512 hazard ratio 1.12 (1.06, 1.18) P less than 0.001; P for interaction 0.336. DBP low less than 79 millimeter of mercury N 445 hazard ratio 1.06 (0.96, 1.18) P 0.223; DBP high greater than or equal to 79 millimeter of mercury N 489 hazard ratio 1.18 (1.11, 1.25) P less than 0.001; P for interaction 0.072. SBP low less than 138 millimeter of mercury N 467 hazard ratio 1.10 (1.01, 1.20) P 0.021; SBP high greater than or equal to 138 millimeter of mercury N 467 hazard ratio 1.17 (1.10, 1.25) P less than 0.001; P for interaction 0.201. Duration low less than 20 years N 451 hazard ratio 1.21 (1.14, 1.28) P less than 0.001; Duration high greater than or equal to 20 years N 482 hazard ratio 1.17 (1.09, 1.26) P less than 0.001; P for interaction 0.208. Drinking history no N 717 hazard ratio 1.15 (1.09, 1.22) P less than 0.001; yes N 217 hazard ratio 1.15 (1.04, 1.28) P 0.007; P for interaction 0.976. Hb low less than 140 gram per liter N 464 hazard ratio 1.19 (1.10, 1.29) P less than 0.001; Hb high greater than or equal to 140 gram per liter N 470 hazard ratio 1.11 (1.05, 1.19) P less than 0.001; P for interaction 0.216. PLT low less than 211 times 10 to the power of 9 per liter N 459 hazard ratio 1.14 (1.05, 1.24) P 0.001; PLT high greater than or equal to 211 times 10 to the power of 9 per liter N 475 hazard ratio 1.15 (1.08, 1.22) P less than 0.001; P for interaction 0.847. AST low less than 20 unit per liter N 442 hazard ratio 1.09 (0.98, 1.20) P 0.106; AST high greater than or equal to 20 unit per liter N 492 hazard ratio 1.17 (1.10, 1.24) P less than 0.001; P for interaction 0.237. TG low less than 1.61 millimole per liter N 467 hazard ratio 1.09 (1.00, 1.18) P 0.043; TG high greater than or equal to 1.61 millimole per liter N 467 hazard ratio 1.19 (1.12, 1.27) P less than 0.001; P for interaction 0.055. HbA1c low less than 9.00 percent N 448 hazard ratio 1.19 (1.11, 1.28) P less than 0.001; HbA1c high greater than or equal to 9.00 percent N 486 hazard ratio 1.14 (1.06, 1.22) P less than 0.001; P for interaction 0.326. Antiplatelet drugs no N 482 hazard ratio 1.16 (1.07, 1.26) P less than 0.001; yes N 452 hazard ratio 1.14 (1.07, 1.21) P less than 0.001; P for interaction 0.616.

Subgroup analysis of BMI and incident CMBs. Forest plot shows hazard ratios (HRs) with 95% CIs for the association between per 1 kg/m2 increase in BMI and incident CMBs across subgroups of sex, age, DBP, SBP, duration of diabetes, drinking history, Hb, PLT, ast, TG, HbA1c, and antiplatelet drugs. Cutoffs are median-based for continuous variables. All estimates are from multivariable-adjusted Cox models. P-values for interaction test effect modification; the vertical dashed line indicates no effect (HR=1.0).

Threshold Effect Analysis of BMI and Incident CMBs Risk

RCS analysis revealed a significant nonlinear association between BMI and incident CMBs risk (overall nonlinearity test p < 0.001). Stratified analysis by gender (Table 3) found distinct association patterns (Figure 4): In males (n=533), the BMI risk threshold was 21.10 kg/m2, and when BMI ≥ 21.10, each unit increase in BMI significantly increased CMBs risk by 18% (HR = 1.18, 95% CI: 1.10–1.26, p < 0.001). In females (n=401), the BMI risk threshold was 26.80 kg/m2, and when BMI ≥ 26.80, each unit increase in BMI significantly increased CMBs risk by 25% (HR = 1.25, 95% CI: 1.02–1.54, p = 0.029).

Table 3.

Threshold Effect Analysis of Baseline BMI on the Risk of Incident CMBs Stratified by Sex (Two-Piecewise Linear Regression Model)

Variable Total (n=936) Male (n=534) Female (n=402)
Model 1 Fitting model by standard linear regression 1.18 (1.11–1.25), <0.001 1.23 (1.13–1.34), <0.001 1.12 (1.02–1.24), 0.018
Model 2 Fitting model by two-piecewise linear regression
Threshold (inflection point K, kg/m2) 22.40 21.10 26.80
BMI < K segment effect, HR (95% CI), p 0.76 (0.42, 1.37) 0.041 0.54 (0.33, 0.90), 0.017 0.97 (0.82, 1.15), 0.734
BMI ≥ K segment effect, HR (95% CI), p 1.22 (1.13, 1.31) <0.001 1.18 (1.10, 1.26), <0.001 1.25 (1.02–1.54), 0.029
Difference between the two segment effects, p 0.036 0.003 0.021
P for likelihood test 0.024 0.035 0.048

Notes: The threshold K indicates the inflection point (ie, the turning point) in the relationship between BMI and the risk of new CMBs. The model was adjusted for sex, age, DBP, SBP, duration of diabetes, drinking history, Hb, PLT, AST, TG, HbA1c, and antiplatelet drugs. HR: Hazard ratio per 1 kg/m2 increase in BMI.

Figure 4.

A two line graphs showing body mass index versus log hazard ratio for risk of cerebral microbleeds by sex. Image A shows a line graph for males with a threshold at BMI 21.10 kg/m superscript 2. The x-axis (BMI) ranges from 15 to 40 and the y-axis (Log HR for CMBs risk) ranges from -4 to 4. A vertical line is at x = 21.10. Curve values include (15, -0.2), (20, -0.4), (25, 0.0), (30, 1.0), (35, 2.0), (40, 2.6). Image B shows a line graph for females with a threshold at BMI 26.80 kg/m superscript 2. The x-axis (BMI) ranges from 15 to 35 and the y-axis (Log HR for CMBs risk) ranges from -2.5 to 2.5. A vertical line is at x = 26.80. Curve values include (15, -1.0), (20, -0.6), (25, -0.2), (30, 0.4), (35, 1.2).

Threshold effect analysis of baseline BMI on the risk of incident CMBs stratified by sex. The restricted cubic spline curves illustrate the non-linear association between baseline BMI and the log hazard ratio (log hr) for incident CMBs (p<0.001). The shaded areas represent 95% confidence intervals. The dashed vertical lines indicate the sex-specific risk thresholds identified by the two-piecewise linear regression model (4A, males: 21.10 kg/m2; 4B, females: 26.80 kg/m2).

Abbreviations: BMI, body mass index; CMBs, cerebral microbleeds; hr, hazard ratio.

Discussion

This exploratory study aimed to evaluate the association between baseline BMI and incident CMBs risk in middle-aged and elderly T2DM patients through a prospective cohort design. During a median follow-up of 30.00 months, we found that baseline obesity (BMI ≥ 28 kg/m2) was associated with incident CMBs risk, with risks 3.52-fold (unadjusted) or 2.58-fold (adjusted) higher than normal weight individuals. In exploratory analyses, we observed a nonlinear association between BMI and CMBs risk with potential gender heterogeneity: the exploratory inflection point for males (21.10 kg/m2) was lower than for females (26.80 kg/m2). Given the exploratory nature of the threshold analysis, these findings should be considered hypothesis-generating rather than definitive.

This study confirms the positive association between obesity and CMBs risk, consistent with most cross-sectional studies.7,8,13 However, compared to cross-sectional designs, prospective studies offer key advantages: they establish temporal precedence between exposure (baseline BMI) and outcome (incident CMBs), providing stronger evidence for causal inference. Our finding that each 1 kg/m2 increase in baseline BMI was associated with a 15% increased risk of future CMBs is comparable to effect sizes reported in general populations,9 suggesting that diabetic patients may not be uniquely susceptible to obesity-related cerebrovascular damage as previously hypothesized. Furthermore, our study found the nonlinear threshold relationship and gender differences between BMI and CMBs risk. After adjusting for potential confounders, including age, sex, blood pressure, and specifically duration of diabetes, baseline obesity remained an independent predictor of incident CMBs. This adjustment ensures that the observed gender-specific BMI thresholds are not biased by differences in the chronicity of the diabetic state among participants. The mechanisms linking obesity to cerebral small vessel damage may involve multiple pathways: (1) Metabolic disturbances and endothelial dysfunction: insulin resistance and dyslipidemia reduce nitric oxide bioavailability and increase vascular permeability;14,15 (2) Chronic inflammation and oxidative stress: visceral adipose tissue releases pro-inflammatory cytokines that disrupt blood-brain barrier integrity;16 and (3) Hemodynamic stress: obesity-related hypertension increases cerebral small artery wall stress and promotes hypercoagulability.17 The “threshold effect” may reflect a cumulative process whereby compensatory mechanisms (eg, anti-inflammatory, antioxidant defenses) become overwhelmed once fat accumulation exceeds a critical point, leading to accelerated CMBs risk. Furthermore, our stratified analysis revealed a stronger association between BMI and deep CMBs than lobar CMBs. This differential effect is biologically plausible, as deep CMBs are primarily attributed to hypertensive arteriolosclerosis, whereas lobar CMBs arise from cerebral amyloid angiopathy (CAA).18 Obesity may promote deep CMBs through hemodynamic stress and hypertension-related small vessel damage, while its effect on CAA-related lobar CMBs appears limited, supporting that obesity predominantly drives microvascular injury via the hypertensive vasculopathy pathway.

The most striking finding of this study is the significant gender-specific threshold. The male risk threshold (21.10 kg/m2) lies at the upper limit of the normal weight range, while the female threshold (26.80 kg/m2) is already within the overweight range. This difference can be understood from several perspectives: (1) Protective effects of sex hormones: Premenopausal estrogen has powerful vasoprotective effects, including suppressing vascular inflammation, improving endothelial function, promoting vasodilation, and inhibiting vascular smooth muscle cell proliferation.19 Given that the mean age of female participants was 63 years, most were postmenopausal; however, residual or cumulative protective effects of estrogen cannot be entirely excluded. Future studies with direct measurements of menopausal status and sex hormone levels are needed to disentangle these effects. (2) Differences in body fat distribution: Men tend toward central (visceral) obesity, while women often have peripheral (subcutaneous) obesity. Visceral fat is more metabolically active, secretes more inflammatory factors, and is more closely related to metabolic syndrome and cardiovascular risk.20 Thus, even at the same BMI, men may experience greater metabolic and vascular damage pressure. (3) Differences in muscle mass: At the same BMI, men typically have higher muscle content. Muscle is an important metabolic and endocrine organ, helping maintain metabolic health. Lower muscle mass in women might suggest higher risk of metabolic abnormalities at lower BMI, but our results show the opposite, suggesting sex hormone and fat distribution differences may be more explanatory. Future research should directly measure body fat distribution and sex hormone levels to test these hypotheses.

This study’s findings have important translational value: 1. Precise Risk Stratification: Clinicians should consider BMI as a key indicator when assessing cerebral small vessel disease risk in diabetic patients, using gender-specific cutoffs (male: 21.10 kg/m2, female: 26.80 kg/m2) for stratification. Patients exceeding these thresholds should be considered high-risk for CMBs. 2. Individualized Weight Management Goals: To prevent cerebral small vessel complications, ideal targets for male diabetic patients should be maintaining BMI below 22 kg/m2; for females, targets can be set below 27 kg/m2. This is more personalized and feasible than the generic “below 24 kg/m2” target. 3. Integrated Screening Strategy: For diabetic patients with elevated BMI, especially those with hypertension, long disease duration, or poor glycemic control, consider regular (eg, every 2–3 years) head SWI-MRI screening to detect CMBs early and adjust treatment strategies promptly.

Limitations

Several limitations should be acknowledged when interpreting our findings. First, this was a single-center study, which may limit the generalizability of the results. Second, we lacked direct measurements of body fat distribution, visceral fat area, or sex hormone levels, which constrains in-depth exploration of the underlying mechanisms. Third, dynamic BMI measurements were not available for all participants during follow-up, preventing analysis of how changes in BMI might influence risk. Fourth, although the SWI sequence exhibits high sensitivity for detecting CMBs, some very small lesions could still have been missed. Fifths, all participants were hospital-recruited, which may introduce Berkson’s bias and potentially overestimate the BMI–CMBs association. Sixth, we lacked data on waist circumference, waist-hip ratio, and detailed diabetic complications, which may serve as important confounders. Seventh, the follow-up duration (median 30 months) is relatively short compared to typical cerebrovascular outcome studies, and may not capture CMBs that develop over longer time horizons. Finally, despite adjustment for multiple covariates, residual confounding from unmeasured factors-such as diet quality, physical activity, or sleep apnea-cannot be entirely ruled out. Future studies incorporating multi-center designs, longitudinal body composition assessments, and comprehensive lifestyle evaluations are warranted to validate and extend our findings.

Conclusion

This prospective cohort study of middle-aged and elderly T2DM patients found that higher baseline BMI was associated with incident CMBs, with a nonlinear threshold association and significant gender differences, suggesting a “critical point” effect of obesity on cerebral small vessel damage. Male BMI > 21.10 kg/m2 and female BMI > 26.80 kg/m2 represent critical points for significantly elevated CMBs risk. This study highlights the importance of implementing gender-differentiated weight control strategies in diabetes management and provides specific BMI intervention targets for preventing cerebral small vessel disease.

Funding Statement

This work was supported by the China Stroke Association Whole Course Management of Cerebrovascular Disease Sailing Fund, 202001 (SDZ).

Data Sharing Statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Author Contributions

Jian Yuan: Conceptualization, Data curation, Writing – original draft. Yun Deng: Data curation, Formal analysis, Writing – original draft. Xingyu Chen: Data curation, Writing – original draft. Linrui Wang: Formal analysis, Writing – review & editing. Feihong Chen: Formal analysis. Ning Rong: Formal analysis, Writing – review & editing, Supervision. Desheng Zhu: Conceptualization, Supervision, Writing – review & editing, Funding acquisition. All authors took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

All authors declare no conflicts of interest.

References

  • 1.Greenberg SM, Vernooij MW, Cordonnier C, et al. Cerebral microbleeds: a guide to detection and interpretation. The Lancet Neurol. 2009;8(2):165–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Akoudad S, Wolters FJ, Viswanathan A, et al. Association of cerebral microbleeds with cognitive decline and dementia. JAMA Neurol. 2016;73(8):934–943. doi: 10.1001/jamaneurol.2016.1017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Charidimou A, Imaizumi T, Moulin S, et al. Brain hemorrhage recurrence, small vessel disease type, and cerebral microbleeds: a meta-analysis. Neurology. 2017;89(8):820–829. doi: 10.1212/WNL.0000000000004259 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Poels MM, Ikram MA, van der Lugt A, et al. Cerebral microbleeds are associated with worse cognitive function: the Rotterdam Scan Study. Neurology. 2012;78(5):326–333. doi: 10.1212/WNL.0b013e3182452928 [DOI] [PubMed] [Google Scholar]
  • 5.Flegal KM, Kit BK, Orpana H, Graubard BI. Association of all-cause mortality with overweight and obesity using standard body mass index categories: a systematic review and meta-analysis. JAMA. 2013;309(1):71–82. doi: 10.1001/jama.2012.113905 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Lavie CJ, Milani RV, Ventura HO. Obesity and cardiovascular disease: risk factor, paradox, and impact of weight loss. J Am Coll Cardiol. 2009;53(21):1925–1932. doi: 10.1016/j.jacc.2008.12.068 [DOI] [PubMed] [Google Scholar]
  • 7.Henskens LH, van Oostenbrugge RJ, Kroon AA, et al. Brain microbleeds are associated with ambulatory blood pressure levels in a hypertensive population. Hypertension. 2008;51(1):62–68. [DOI] [PubMed] [Google Scholar]
  • 8.Miwa K, Tanaka M, Okazaki S, et al. Relations of blood inflammatory marker levels with cerebral microbleeds. Stroke. 2011;42(11):3202–3206. doi: 10.1161/STROKEAHA.111.621193 [DOI] [PubMed] [Google Scholar]
  • 9.Jeerakathil T, Wolf PA, Beiser A, et al. Cerebral microbleeds: prevalence and associations with cardiovascular risk factors in the Framingham Study. Stroke. 2004;35(8):1831–1835. doi: 10.1161/01.STR.0000131809.35202.1b [DOI] [PubMed] [Google Scholar]
  • 10.World Health Organization. Classification of Diabetes Mellitus. Geneva: World Health Organization; 2019. [Google Scholar]
  • 11.Chen C, Lu FC. Department of Disease Control Ministry of Health PRC. The guidelines for prevention and control of overweight and obesity in Chinese adults. Biomed Environ Sci. 2004;17 Suppl:1–36. [PubMed] [Google Scholar]
  • 12.Consultation WHOE. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet. 2004;363(9403):157–163. doi: 10.1016/S0140-6736(03)15268-3 [DOI] [PubMed] [Google Scholar]
  • 13.Yamashiro K, Tanaka R, Tanaka Y, et al. Visceral fat accumulation is associated with cerebral small vessel disease. Eur J Neurol. 2014;21(4):667–673. [DOI] [PubMed] [Google Scholar]
  • 14.Guzik TJ, Marvar PJ, Czesnikiewicz-Guzik M, Korbut R. Perivascular adipose tissue as a messenger of the brain-vessel axis: role in vascular inflammation and dysfunction. J Physiol Pharmacol. 2007;58(4):591–610. [PubMed] [Google Scholar]
  • 15.Vanhoutte PM, Shimokawa H, Tang EH, Feletou M. Endothelial dysfunction and vascular disease. Acta Physiol. 2009;196(2):193–222. doi: 10.1111/j.1748-1716.2009.01964.x [DOI] [PubMed] [Google Scholar]
  • 16.Wang Z, Hoy WE. Albuminuria as a marker of the risk of developing type 2 diabetes in non-diabetic Aboriginal Australians. Int J Epidemiol. 2006;35(5):1331–1335. doi: 10.1093/ije/dyl115 [DOI] [PubMed] [Google Scholar]
  • 17.Caso V, Paciaroni M, Agnelli G, et al. Obesity paradox and stroke: a review of the literature. Stroke. 2011;42(1):e30–e33. [Google Scholar]
  • 18.Haller S, Vernooij MW, Kuijer JPA, Larsson EM, Jäger HR, Barkhof F. Cerebral microbleeds: imaging and clinical significance. Radiology. 2018;287(1):11–28. doi: 10.1148/radiol.2018170803 [DOI] [PubMed] [Google Scholar]
  • 19.Mendelsohn ME, Karas RH. The protective effects of estrogen on the cardiovascular system. N Engl J Med. 1999;340(23):1801–1811. doi: 10.1056/NEJM199906103402306 [DOI] [PubMed] [Google Scholar]
  • 20.Després JP, Lemieux I. Abdominal obesity and metabolic syndrome. Nature. 2006;444(7121):881–887. doi: 10.1038/nature05488 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.


Articles from Diabetes, Metabolic Syndrome and Obesity are provided here courtesy of Dove Press

RESOURCES